clock menu more-arrow no yes mobile

Filed under:

A Very Serious Review of an AI-Generated Christmas Song

Our own take on the Canadian newcomer’s seasonal release

(Getty Images/Ringer illustration)
(Getty Images/Ringer illustration)

Many great things are expected of artificial intelligence: machines that tell us definitively whether or not a hot dog is a sandwich, our very own Westworld-like theme park, and a strict but benevolent robot ruling class. But we’re not quite there yet, and in the meantime AI is putzing around, mimicking Friends characters and becoming the world champion of Go.

Now, just in time for the holidays, one particularly festive bot has taken on Christmas music. Per a report in The Guardian, a neural network at the University of Toronto recently composed and sang an experimental 57-second holiday ditty, based on a photo of a Christmas tree. Critics have not been kind. “It will not, if there is any certainty left in the world, top the charts this Christmas,” The Guardian jeered. Meanwhile, bots around the nation are calling it “01010101-DELETE.” Here is our own take on the Canadian newcomer’s seasonal release:

Though University of Toronto Neural Network is new to the scene, it remains impressively meticulous in its artistic approach. The network spent 100 hours processing online music, 50 hours taking in internet song lyrics, and — most fascinatingly — two hours analyzing footage from Just Dance video games, before it even began the songwriting process, an activity that its producer, Ph.D. student Hang Chu, has warmly referred to as “neural karaoke” in interviews. While NN’s deference to its musical predecessors is respectable, it’s not necessarily original. You’ll recall that the AI known as Flow Machines underwent a similar education during the making of its eerily vacant Beatles-esque single “Daddy’s Car,” solidifying itself as the creator of the genre (even if it did embarrassingly lean on a human lyric writer for the release).

Unlike Flow Machines, NN’s “Neural Story Singing Christmas” is completely human-free. Reportedly inspired by a single photograph of a Christmas tree, the holiday tune begins abruptly, as an elf-like voice serenades against a playful piano tune. At first it sounds like something you’d subconsciously tune out while rifling through a panty drawer at the Victoria’s Secret Black Friday sale. But with a closer listen, it becomes clear NN is only invoking the framework of traditional holiday music in order to turn it on its head.

A couple of seconds in, the song strikes a sinister piano key and its lyrics creep into anxious territory: “Lots to decorate the room / The Christmas tree is filled with flowers / I swear it’s Christmas Eve / I hope that’s what you say.” It seems that despite the warm and joyful occasion, Neural Network still feels as though it’s in a strange, faraway land, where ornaments are forsaken for flowers, and it’s not even sure whether it’s actually Christmas — perhaps the experience of a sentient being who is forced to kowtow to the simplistic cultural traditions of humans.

A lesser ear may take NN’s unique lyrical flare — the tendency to sing as if it were in a Garth and Kat SNL sketch — to be a sign of immaturity as a musician. But I would argue that this is an intentional flourish, meant to make the listener uncomfortable and help them experience the same creative limitations that Neural Network feels within the confines of the human Christmas music canon. This poetic imprisonment is underlined as the piano tune builds to a dramatic crescendo and NN sings: “I’ve always been there for the rest of our lives / A hundred and a half hour ago.” How boring “Jingle Bells” must seem to a bot that knows no limits to its intelligence.

Within a millisecond, however, University of Toronto Neural Network returns to the Christmas-song cheese, summarizing its celebration the same way Ariana Grande might, sans enthusiasm: “I can hear the music coming from the hall / A fairy tale / A Christmas tree.” It’s as if Neural Network knows its suffering will go without notice, and it’s playing the game … for now.